Artificial Analysis Coding Agent Benchmarks
Here at Artificial Analysis, we are excited to have answered the question on every developer's mind: which model and harness actually work best together?
As our new Coding Agent Benchmarks show, the combination matters more than people expect. Across the agents we tested, cost per task varied by more than 30x and speed by more than 7x. The next phase of progress for coding agents will come from a mix of model and harness improvement.
Come and join the Artificial Analysis for an evening of talks and discussions with the team and special guests from Cursor, Cognition, and Kernel Labs, digging into coding agent benchmarks. The room will be full of researchers, agent builders, and AI engineers, and plenty of time to argue over food and drinks.
Agenda:
6:00 PM: Networking, food, and drinks
6:30 PM: Lightning talks
7:30 PM: Panel (Speakers to be announced)
8:00 PM: Networking, food, and drinks
8:30 PM: Event ends
Registrations are reviewed to keep the room focused on people building in AI. Space has limited availability, so register early! We’re looking forward to seeing you at the event.
Photography disclaimer: By attending this event, you acknowledge that photography may take place; if you do not consent, please contact the event host.